Non-overlapping epitopes on the gHgL-gp42 complex for the rational design of a triple-antibody cocktail against EBV infection

Summary

Epstein-Barr virus (EBV) is closely associated with cancer, multiple sclerosis, and post-acute coronavirus disease 2019 (COVID-19) sequelae. There are currently no approved therapeutics or vaccines against EBV. It is noteworthy that combining multiple EBV glycoproteins can elicit potent neutralizing antibodies (nAbs) against viral infection, suggesting possible synergistic effects. Here, we characterize three nAbs (anti-gp42 5E3, anti-gHgL 6H2, and anti-gHgL 10E4) targeting different glycoproteins of the gHgL-gp42 complex. Two antibody cocktails synergistically neutralize infection in B cells (5E3+6H2+10E4) and epithelial cells (6H2+10E4) in vitro. Moreover, 5E3 alone and the 5E3+6H2+10E4 cocktail confer potent in vivo protection against lethal EBV challenge in humanized mice. The cryo-EM structure of a heptatomic gHgL-gp42 immune complex reveals non-overlapping epitopes of 5E3, 6H2, and 10E4 on the gHgL-gp42 complex. Structural and functional analyses highlight different neutralization mechanisms for each of the three nAbs. In summary, our results provide insight for the rational design of therapeutics or vaccines against EBV infection.

Graphical abstract

Highlights

• •nAbs targeting different glycoproteins synergistically neutralize EBV infection
• •gp42 nAb and its triple-antibody cocktail confer potent protection in humanized mice
• •Structure reveals non-overlapping epitopes and various neutralizing mechanisms

Epstein-Barr virus is linked to multiple diseases, and there is no approved vaccine. Hong et al. characterize antibodies targeting different components of the gHgL-gp42 complex, demonstrating in vitro synergistic neutralization and potent in vivo protection in a triple-antibody format, providing insights for the rational design of therapeutics or vaccines.

Introduction

Epstein-Barr virus (EBV), the first identified oncogenic virus, infects more than 95% of adults worldwide.¹ EBV is strongly linked with multiple types of cancer, including Hodgkin’s lymphoma, Burkitt’s lymphoma, large B cell lymphoma, post-transplantation lymphoproliferative disease, natural killer/T cell lymphoma, nasopharyngeal carcinoma (NPC) and gastric carcinoma.² Each year, EBV is responsible for 200,000 new cases of malignancies and 140,000 deaths.³ In addition, EBV is also the causative agent of infectious mononucleosis (IM),¹ and a recent longitudinal analysis identified EBV as the leading cause of multiple sclerosis.⁴ Besides, EBV reactivation may be the cause of post-acute sequelae of coronavirus disease 2019 (COVID-19);⁵ COVID-19 patients with EBV reactivation show more severe symptoms and require more immune treatment⁶ with a worse prognosis and increased mortality rates.⁷ Therefore, the prevention of EBV infection or reactivation may improve survival among patients with COVID-19. However, to date, no prophylactic vaccine or effective therapy has been approved against EBV infection or EBV-associated diseases.⁸

EBV primarily infects two types of host cells, B cells and epithelial cells,^9,10 through a complex, multistep process involving several viral glycoproteins.^11,12 At least five reported EBV glycoproteins (gp350/220, gH, gL, gp42, and gB) are involved in the viral entry process,¹³ of which four (gH, gL, gp42, and gB) and three (gH, gL, and gB) are indispensable for membrane fusion with B cells and epithelial cells, respectively.¹⁴ The gHgL heterodimer, comprising membrane anchor gH and soluble gL,¹⁵ is an important component of the EBV entry complex and, together with gB, forms the core fusion machinery that participates in EBV infection within both host cell types.¹⁶ During entry into epithelial cells, gHgL binds directly to a range of receptors, including Ephrin type A receptor 2 (EphA2) and integrins αvβ5, αvβ6, and/or αvβ8.^17,18 Apart from gHgL, gp42 is also a key entry component for EBV infection and, in particular, a viral tropism determinant that promotes B cell infection while inhibiting epithelial cell infection.¹⁹ During B cell infection, gp42 binds tightly to heterodimeric gHgL, forming a stable heterotrimeric gHgL-gp42 complex, and interacts with the host receptor human leukocyte antigen class II (HLA-II), promoting EBV entry into B cells.²⁰

EBV envelope glycoproteins are promising targets for neutralizing antibodies (nAbs), and a number of nAbs against different antigens have been reported.²¹ Characterization of the murine nAb 72A1 led to the identification of gp350, the major glycoprotein of EBV,²² with 72A1 able to effectively neutralize viral infection of B cells, but not epithelial cells, by blocking gp350 binding to the B cell receptor CD21 and CD35.^23,24 Many other anti-gp350 nAbs can also inhibit B cell infection, whereas none can yet neutralize epithelial cell infection.^25,26 AMMO5 is an anti-gB nAb that shows biased neutralization against B cell infection.²⁷ Recently, we reported two rabbit-derived anti-gB nAbs, 3A3 and 3A5, each targeting a discrete vulnerable site but both showing dual-tropic inhibition of B cell and epithelial cell infection.²⁸ gHgL is the core component of the fusion machinery and is thought to be a promising prophylactic EBV vaccine candidate.²⁹ Hence, there has been increasing interest in isolating anti-gHgL nAbs. The gHgL-specific murine nAbs E1D1, CL40, and CL59 are capable of neutralizing epithelial cell infection but not B cell infection.^30,31 We have also recently isolated one gH-specific nAb 6H2 that showed complete and partial neutralizing activities in epithelial cells and B cells, respectively.³² Three human anti-gHgL nAbs, AMMO1, 1D8, and 769B10, have also been reported to show dual-tropic inhibition of both cell types.^27,33,34 Furthermore, AMMO1, 1D8, and 6H2 have shown promising protection efficacy against EBV infection in a humanized mouse model.^27,32,33

gp42 is another EBV entry component that is essential for B cell infection; however, the humoral immune response to gp42 is poorly understood.³⁵ Indeed, few gp42-specific nAbs have been reported,^35,36,37 and the structure and neutralizing mechanisms of anti-gp42 nAbs remain largely unknown. A recent study of vaccinated monkeys revealed that more potent fusion-inhibitory antibodies could be induced via the combination of gp42 and gHgL as opposed to gHgL alone, suggesting a potential synergistic effect between anti-gp42 and anti-gHgL nAbs.³⁴ Nevertheless, little attention has been paid to investigating the synergistic mechanism of nAbs targeting different EBV glycoproteins.

To better characterize gp42-specific nAbs and understand the potential synergistic effects between anti-gp42 and anti-gHgL nAbs, here we analyzed the activities and neutralizing potencies of three nAbs—5E3, 10E4, and the previously reported 6H2³²—majorly targeting gp42, gL, and gH, respectively. 5E3 showed comparable B cell-neutralizing potency with the previously described anti-gHgL nAb AMMO1.²⁷ The combination of three (5E3+6H2+10E4) and two (6H2+10E4) antibodies showed significant synergistic neutralization in B cells and epithelial cells, respectively. Moreover, 5E3 mAb and a 5E3-based triple-antibody cocktail conferred potent protective activities against lethal challenge with EBV infection in a humanized mouse model. Using cryoelectron microscopy (cryo-EM), we resolved the high-resolution structure of the immune complex of gHgL-gp42 bound simultaneously by 5E3, 6H2, 10E4 and another non-neutralizing anti-gp42 monoclonal antibody (mAb), 3E8. This structure of a seven-component immune complex allowed us to define four specific determinants on gHgL-gp42 that distinguish this binding from that of other reported nAbs and, together with our functional study, revealed the diverse neutralizing mechanisms of nAbs 5E3, 6H2, and 10E4. The synergistic effects between anti-gp42 and anti-gHgL nAbs suggest the importance of combining nAbs targeting different glycoproteins to combat EBV infection and EBV-induced disease.

Results

Isolation and characterization of nAbs targeting EBV gp42 and gHgL

The soluble form of gp42 and gHgL were constructed and expressed as described previously (Figure S1A).^32,36 Purified gp42 and gHgL migrated as single bands with molecular weights of ∼42 and ∼110 kDa, respectively, with corresponding retention times of 17.00 min and 14.30 min, respectively, indicative of their high purity (Figures S1B–S1D). Recombinant gp42 and gHgL were recognized by human sera, suggesting that they were suitably folded and maintained the native antigenicity (Figure S1E). gp42 and gHgL were used as immunogens for subsequent rabbit immunization. Allophycocyanin (APC)-conjugated gp42 and gHgL served as bait to isolate antigen-specific memory B cells via the rabbit mAb platform (Figures S1F and S1G).²⁸ Finally, two anti-gp42 rabbit mAbs, 3E8 and 5E3, and one anti-gHgL mAb, 10E4, were obtained.

To further analyze and compare the properties of antibodies targeting different glycoproteins of EBV, two previously reported anti-gHgL nAbs, 6H2³² and AMMO1,²⁷ were included in our study. To reduce the potential influence of Fc effector function, the constant regions of 3E8, 5E3, 6H2, and 10E4 were replaced with the human immunoglobulin G1 (IgG1) constant region. The anti-HIV-1 mAb VRC01³⁸ was selected as a negative control. Through antibody binding competition assays, we found that 3E8 and 5E3 could bind to gp42 together (Figure S2A), with no competition observed from 6H2, 10E4, or AMMO1 (Figure S2B). In addition, all mAbs could bind simultaneously to the gHgL-gp42 complex, suggesting that they recognized different epitopes on the gHgL-gp42 complex without steric hindrance (Figure S2C).

We first evaluated the binding of the aforementioned mAbs against recombinant glycoproteins using surface plasmon resonance (SPR). The dissociation constant (K_D) values measured by SPR showed subnanomolar binding affinities for 3E8, 5E3, and AMMO1 (3E8, 0.42 nM; 5E3, 0.39 nM; AMMO1, 0.43 nM), and nanomolar binding affinities for 6H2 and 10E4 (6H2, 6.83 nM; 10E4, 1.93 nM) (Figures 1A and S2D–S2H). Next, we investigated the binding activities of the mAbs with the virion-encoded gp42 or gHgL expressed on the cell surface. Here again, with the exception of VRC01, all could specifically bind to EBV-positive CNE2 cells (induced for EBV lytic production), indicating that these EBV-specific mAbs recognized the native gp42 or gHgL (Figure 1B).

Figure 1.

Binding and neutralizing efficacies of anti-gp42 and anti-gHgL mAbs

(A) The binding affinities of gp42-and gHgL-specific mAbs were measured by SPR assay, performed using sensor chip protein A on a BIAcore 8K. The sensorgram curve for each mAb is shown in Figures S2D–S2H.

(B) Binding activities of mAbs to cell-surface-expressed gp42 or gHgL on EBV-positive CNE2 cells using flow cytometry. AMMO1 and VRC01 were used as positive and negative controls, respectively. Alexa Fluor 647 (AF647) anti-human IgG secondary antibody was used for detection.

(C and D) Neutralizing activities of the above antibody and antibody cocktails against EBV infection of Akata B cells (C) and HNE1 epithelial cells (D). The total concentration in the antibody cocktail was equal to that for the single mAb, with the antibody cocktail comprising an equal molar ratio of each mAb. The half-maximal inhibitory concentration (IC₅₀) was calculated and is indicated.

Data from one experimental replicate are shown in (A) and (B). Data are shown as the mean ± standard error of the mean (SEM) from two replicates in (C) and (D). See also Figures S1 and S2.

We next evaluated the abilities of mAbs to neutralize EBV infection in B cells and epithelial cells. Except for 3E8, all of the other EBV-specific mAbs showed effective potency in blocking B cell infection (Figure 1C). Anti-gp42 nAb 5E3 exhibited excellent and complete neutralization of B cell infection, with a half-maximal inhibitory concentration (IC₅₀) of 0.20 μg/mL, which was comparable with that of AMMO1 (0.30 μg/mL). 6H2 and 10E4 showed less potent B cell neutralization, with IC₅₀ values of 7.01 μg/mL and 15.15 μg/mL, respectively. We also assessed the synergistic potency of anti-gp42 nAb 5E3, anti-gHgL nAb 6H2, and anti-gHgL nAb 10E4. In the B cell infection assay, the combination of 5E3+6H2 (molar ratio 1:1, IC₅₀ = 0.13 μg/mL), 5E3+10E4 (molar ratio 1:1, IC₅₀ = 0.18 μg/mL), or 6H2+10E4 (molar ratio 1:1, IC₅₀ = 1.41 μg/mL) showed enhanced neutralization compared with individual antibodies (Figure S2I). In addition, the triple-antibody cocktail (molar ratio 1:1:1) showed potent neutralization with an IC₅₀ value of 0.08 μg/mL (Figure 1C), which surpasses the neutralization potency of any individual nAb and bivalent antibody mixture, clearly suggesting potential synergistic neutralization activity of three nAbs.

As for the epithelial cell infection assay, anti-gp42 nAbs 3E8 and 5E3 showed no neutralizing effects, as expected (Figure 1D), because gp42 does not participate in epithelial cell infection. In contrast, the neutralization of 6H2 and 10E4 was significantly better in epithelial cells than in B cells, with IC₅₀ values of 0.13 and 0.74 μg/mL, respectively (Figure 1D). In addition, we observed a synergistic neutralizing effect by combining 6H2 and 10E4 (0.07 μg/mL) that was comparable with that of AMMO1 (0.05 μg/mL) (Figure 1D). From these collective results, we see that 5E3+6H2+10E4 and 6H2+10E4 combinations offer synergistic neutralization of infection in B cells and epithelial cells, respectively.

5E3 and the 5E3-based cocktail protect humanized mice from lethal EBV challenge

We previously confirmed the in vivo protective effect of 6H2 using a humanized mouse model established by engrafting non-obese diabetic (NOD)-Prkdc^null IL2Rγ^null (NPI) mice with CD34⁺ human hematopoietic stem cells.³² Here we used the same mouse model to evaluate the in vivo protective potential of the mono-antibody 5E3, 10E4, and the antibody cocktail 5E3+10E4+6H2. The in vivo viral challenge and antibody administration are shown in Figure 2A. The anti-EBV gHgL nAb AMMO1²⁷ and the anti-HIV mAb VRC01³⁸ were selected as positive and negative controls, respectively. Briefly, humanized mice received 400 μg (∼20 mg/kg) of mono-antibody or the antibody-cocktail (1:1:1 mixture) through intraperitoneal (i.p.) injection. 24 h post antibody treatment, mice were challenged with a lethal intravenous (i.v.) injection of Akata-EBV, equivalent to ∼25,000 green Raji units (GRUs). The mice in the uninfected control group did not receive antibody or viral injection. Mice were monitored over the next 10 weeks for survival, body weight changes, EBV DNA copy numbers, and immunophenotype analysis. Mice were euthanized at week 10 or earlier when clinically ill (e.g., more than 20% body weight loss), and tissue pathology was evaluated.

Figure 2.

5E3 and the 5E3+6H2+10E4 cocktail provide potent protection against lethal EBV challenge in humanized mice

(A) Experimental timeline for CD34⁺ hematopoietic stem cell (HSC) engraftment, antibody infusion, EBV challenge, and monitoring for various biological and clinical outcomes. A total of 400 μg of 5E3 (n = 6), 10E4 (n = 6), 5E3+6H2+10E4 cocktail (n = 6), AMMO1 (positive control, n = 5), and VRC01 (negative control, n = 5) was administered to NPI mice via intraperitoneal (i.p.) injection 24 h before intravenous (i.v.) challenge with Akata-EBV (25,000 GRUs). Uninfected control mice (n = 3) did not receive an injection of antibody or virus. Body weight, animal survival, EBV DNA copy number in the peripheral blood, and immunophenotype were recorded weekly for 10 weeks. Humanized mice were euthanized at week 10, and spleens were collected for further analysis.

(B–D) Body weight (B), survival (C), and EBV DNA copy number in the peripheral blood samples (D) of mice were monitored weekly. Each line in (D) represents an individual mouse; the limit of detection (10 copies/μL) is indicated by a dashed line. Body weight was converted to percentage.

(E–I) The percent changes in lymphocyte proportions of hCD45⁺ (E), hCD19⁺ (F), hCD3⁺ (G), hCD8⁺ (H), and hCD4⁺ (I) cells in the peripheral blood were analyzed using flow cytometry.

All data are from one experimental replicate and presented as the mean ± SEM. Statistical analyses were performed using one-way ANOVA, comparing each group of data with that of the VRC01 group. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001; ns, not significant. 5E3, 10E4, 5E3+6H2+10E4 combination, AMMO1, VRC01, and uninfected control are colored green, purple, deep pink, gray, brown, and black, respectively. See also Figure S3.

VRC01-treated mice started to significantly lose body weight 4 weeks post challenge, and all mice died 7 weeks post challenge (Figures 2B and 2C). Mice in all of the other mAb-treated groups maintained relatively stable body weight (Figure 2B). Mice receiving 5E3 or the antibody cocktail and uninfected controls showed 100% survival after challenge (Figures 2B and 2C), whereas only 50% and 80% of 10E4- and AMMO1-treated mice survived to 10 weeks (Figure 2C). EBV viremia (>10 copies/μL) was observed in all VRC01-treated mice and three (50%) 10E4-treated mice (Figure 2D). In VRC01-treated mice, EBV DNA copy number increased rapidly from 2 weeks after challenge and was maintained at a high level (>1,000 copies/μL) until death. In contrast, none of the mice in the 5E3, 5E3+6H2+10E4 cocktail, AMMO1, or uninfected control groups had blood EBV DNA copy numbers greater than 10 copies/μL, suggesting that infection was reduced or blocked to undetectable levels during the observation period (Figure 2D). Consistent with the viral load in peripheral blood draws, VRC01-treated mice showed the highest EBV DNA copy number in the spleen, at a 10-fold higher level than that measured for the 10E4-treated group and more than 1,000-fold higher than that for the 5E3-, antibody cocktail-, and AMMO1-treated groups (Figure S3A). Altogether, these results exemplify the promising therapeutic potencies of 5E3 mAb and the 5E3-based triple-antibody cocktail.

To evaluate the engraftment efficiency of the humanized mice model, the percentage of human leukocytes (hCD45⁺) was analyzed. We identified similar proportions of hCD45⁺ lymphocytes in the peripheral blood draws of mice over the 10-week observation period (Figure 2E), suggesting the successful reconstruction and stable sustainability of the human immune system. Before EBV challenge (week 0), more than 90% of hCD45⁺ lymphocytes in the peripheral blood were hCD19⁺ B cells, the primary target of EBV infection, and less than 1% of cells were identified as hCD3⁺ T cells (Figures 2F and 2G). In humanized mice, B cells are susceptible to viral infection following EBV inoculation, which may be preferentially recognized and eliminated by T cells, primarily hCD8⁺ T cells.³⁹ Mice treated with VRC01 showed a significant (73%) decrease in the proportion of hCD19⁺ B cells from nearly 95% (week 0) to approximately 22% (week 6), whereas the proportion of hCD19⁺ B cells measured for mice in the 10E4 group showed a relatively slower rate of decline from week 0 to week 6 (Figure 2F; 56% decrease over 6 weeks). In contrast, mice treated with 5E3 (26% decrease), antibody cocktail (35% decrease). or AMMO1 (30% decrease) showed much slower rates of decline in the proportion of hCD19⁺ B cells over the 6-week period (week 0 to week 6; Figure 2F). Comparatively, we measured a marked increase in hCD3⁺ T cell levels to more than 80% in VRC01-treated mice, and this was accompanied by a significant increase in the percentage of hCD8⁺ T cells (Figures 2G and 2H). Among the anti-EBV-nAb-treated mice, we noted a relatively slower increase in hCD3⁺ and hCD8⁺ T cell proportions, but 10E4 was relatively inefficient in inhibiting the increase in T cell numbers (Figures 2G and 2H). We observed a minimal change in hCD4⁺ cell numbers in all mice (Figure 2I). Overall, these results confirm that 5E3 and the 5E3-based antibody cocktail can effectively inhibit EBV replication and protect humanized mice from lethal EBV challenge in vivo. Protection from 10E4 is less effective, and this finding is consistent with its weaker neutralizing ability defined in vitro (Figures 1D and 1E).

Viral replication and tissue damage are prevented in the protected animals

We next evaluated the pathological changes of EBV-induced lymphoma at necropsy. Spleens from VRC01-treated mice showed significant morphological changes, with a remarkable enlargement in size, which was much heavier and longer than that from other groups (Figures S3B and S3J). Pale tumors were also observed on the surfaces of spleens from all VRC01-treated and some (three of six) 10E4-treated mice (Figure S3J). In situ hybridization of EBV-encoded RNAs (EBERs) and immunohistochemical (IHC) staining for human CD20 on spleen paraffin sections confirmed that VRC01-treated mice developed typical lymphomas as a result of excessive EBV replication and the outgrowth of EBV-infected B cells (hCD20⁺ and EBER⁺) (Figure S3J). To some extent, we surmise that 10E4-treated mice exhibited modest disease features (small splenic size, lymphomas) because 10E4 can only partially block viral replication in vivo (Figures 2D, S3A, and S3J). On the contrary, spleens from 5E3-, antibody cocktail-, and AMMO1-treated mice showed much less hCD20⁺/EBER⁺ cell infiltration and presented with relatively normal and intact tissue architecture (Figure S3J).

Immune cell population changes observed in the tissues were similar to those observed in the peripheral blood. Spleens of mice treated with different mAbs showed relatively comparable proportions of hCD45⁺ cells (Figure S3C). hCD19⁺ B cells from VRC01-treated and some 10E4-infused mice were significantly depleted compared with that for 5E3-, triple-antibody cocktail-, and AMMO1-treated mice (Figure S3D). However, we identified a remarkable increase in the proportion of highly proliferating memory B cells—hCD19⁺hCD24⁻hCD38^high lymphocytes—in mice in the VRC01-treated group and slight elevation in the 10E4-treated group (Figure S3E), suggesting the outgrowth of EBV-infected B cells; this is consistent with the characteristics of the lymphomas (Figure S3J). Concurrent with the depletion of hCD19⁺ B cells, VRC01- and 10E4-treated mice showed much higher levels of hCD3⁺ and hCD8⁺ T cells in spleens compared with other groups (Figures S3F and S3G). 5E3-, 10E4-, antibody cocktail-, and AMMO1-treated mice had similar hCD4⁺ T cells populations, which were much lower in the VRC01-treated mice (Figure S3H). The proportion of activated hCD8⁺hCD137⁺hCD69⁺ T lymphocytes was significantly higher in VRC01-treated mice (Figure S3I) compared with mice in the other groups, suggesting that the dramatic decrease in hCD19⁺ B cells (Figure S3D) may be attributed to the T cell killing of the EBV-infected B cells by activated CD8⁺ T cells.

Taken together, these data confirm that 5E3 and the 5E3-based cocktail treatment significantly inhibit viral replication in vivo and protect humanized mice from EBV-induced lymphomas. Compared with the incomplete protection of AMMO1, 5E3-based mono- and triple-antibody treatments provide 100% protection against lethal EBV infection challenge. Although 10E4 is less potent when administrated individually, a cocktail containing 5E3 and 6H2 still showed a robust protective effect.

Structural basis for the synergistic neutralization of EBV gHgL-gp42 antibodies

To investigate the structural basis for the in vitro synergistic efficacy of different nAbs targeting the gHgL-gp42 machinery, we prepared gHgL-gp42 in complex with the antibody fragments (Fabs) of the four tested mAbs: 5E3, 3E8, 6H2, and 10E4. High-performance size-exclusion chromatography (HPSEC) analysis showed that four mAbs could bind simultaneously to the gHgL-gp42 complex (Figure S4A). We then determined the cryo-EM structure of the heptatomic immune complex of gHgL-gp42:5E3:3E8:6H2:10E4 to an overall resolution of 3.59 Å (Figures 3A, 3B, and S4B; Table S1). The EBV gHgL-gp42 structure appears as an inverted “T” shape, and each of the four antibodies binds to a distinct epitope: 3E8 and 5E3 bind simultaneously to the apex of gp42, 6H2—consistent with our previous epitope study based on median-resolution cryo-EM structure³²—binds to an epitope spanning the interface between gH D-III and D-IV, and 10E4 binds mainly to gL located within the D-I region (Figures 3A and 3B).

Figure 3.

Composite cryo-EM structures of gHgL and gp42 bound by four antibodies

(A and B) Cryo-EM density maps of the gHgL-gp42:5E3:3E8:6H2:10E4 complex at 3.59-Å resolution. gp42 is colored pink, and gH D-I with gL, D-II, D-III, and D-IV are colored dark khaki, cornflower blue, medium aquamarine, and steel blue, respectively. The densities of 3E8, 5E3, 6H2, and 10E4 are colored orange, cyan, red, and purple, respectively.

(C) Density maps of localized refinement for 3E8- and 5E3-bound gp42 (gp42:5E3:3E8), 6H2-bound gH (gH:6H2), and 10E4-bound gHgL (gHgL:10E4) are resolved at 3.64-Å, 3.52-Å, and 3.54-Å resolution, respectively.

(D) Atomic model of gHgL-gp42:5E3:3E8:6H2:10E4. mAb 3E8, 5E3, 6H2, and 10E4 are presented in cartoon form; gHgL and gp42 are also presented in cartoon form with transparent surface representations. Dark and light coloration is used to differentiate between light and heavy chains, respectively, of each antibody.

(E) Epitope mapping of four antibodies—3E8, 5E3, 6H2, and 10E4—on the surfaces of gp42 (misty rose surface representation), gH (gray surface representation) and gL (beige surface representation). Shared residues between 3E8 and 5E3 epitopes are indicated and colored brown.

(F, G, L, and O) Interaction interfaces of 5E3-gp42 (F), 3E8-gp42 (G), 6H2-gH (L), and 10E4-gHgL (O). Epitope residues are shown in cartoon representation with a transparent surface and labeling.

(H, I, J, K, M, N, P, and Q) Interaction details between gp42 and 5E3 (H and I), gp42 and 3E8 (J and K), gH and 6H2 (M and N), and gHgL and 10E4 (P and Q). Antibodies and targeted antigens are shown in cartoon and surface representations, respectively. Hydrogen bond and salt bridge interactions are shown as yellow and cyan dashed lines, respectively, with involved residues shown as sticks and labeled.

See also Figures S4 and S5, Table S1, and Video S1.

Local resolution analysis revealed relatively high resolution of the main body region of the complex (e.g., gHgL D-II and D-III) but lower resolution of the antibody interfaces, which is caused by the inherent flexibility of the gHgL-gp42 complex (Video S1). To improve the resolution of antibody interfaces, we further performed localized refinement of 3E8- and 5E3-bound gp42 (gp42:5E3:3E8), 6H2-bound gH (gH:6H2), and 10E4-bound gHgL (gHgL:10E4) and obtained localized structures at a resolution of 3.64 Å, 3.52 Å, and 3.54 Å, respectively (Figures 3C and S4B; Table S1). These higher-resolution details allowed us to build the atomic model of the whole gHgL-gp42 and the variable domains of the four bound Fabs (Figures 3D and S4C–S4I).

The four mAbs bind to various epitopes scattered across the surface of the gHgL-gp42 complex (Figure 3E). 5E3 and 3E8 bind apical epitopes on gp42, burying surface areas of about 800 Å² and 835 Å², respectively (Figures 3F and 3G). A total of 18 and 14 residues from gp42 participate in the interaction of 5E3 and 3E8, respectively, of which amino acids (aa) T104, R105, and S221 are involved in the interactions of both antibodies (Figures 3F and 3G). 5E3 mainly uses its long complementarity-determining region 3 of the heavy chain (CDRH3) to meditate its interaction with the loop (aa 104–112) and the helix (aa 147–155) of gp42, where residues A97, R102, D103, Y104, and D106 form 10 hydrogen bonds and many contact interactions with gp42 (Figure 3H). In addition, CDR1 of the light chain (CDRL1) and CDRL3 of 5E3 also form an additional two hydrogen bonds (Figure 3I). The 3E8 interface contains 9 hydrogen bonds, in which CDRL1, CDRL2, CDRL3, CDRH1, and CDRH2 interact with gp42. The light chain of 3E8 mainly interacts with the aa 101–106 loop and the heavy chain with the aa 133–138 loop (Figures 3J and 3K).

6H2 has been determined previously to bind an epitope on gHgL D-IV critical for viral attachment and membrane fusion.³² Here, the high-resolution structure revealed that 6H2 interacts with gHgL through the extension of the framework region of the light chain into the groove between gHgL D-III and D-IV, particularly D-IV, and results in an extremely large buried area of about 1,300 Å² (Figure 3L). The 6H2 epitope consists of up to 24 residues, and the interface comprises a variety of interactions, including a dozen hydrogen bonds and three salt bridge interactions (Figures 3M and 3N). CDRL1 and CDRH3 contact the aa 623–629 loop between the 4β-7 and 4β-8 strands and the aa 648–656 loop between 4α-1 and 4β-9 of gHgL D-IV (Figures 3M and 3N). The fourth antibody, 10E4, binds to D-I of gHgL, which comprises residues of gH and gL, burying a surface area of about 1,090 Å² (Figure 3O). All six CDRs of 10E4 are involved in the interaction and make contact primarily with the Lα-2 helix (aa 104–117) and the short loop (aa 118–120) of gL, with an interaction network of 17 hydrogen bonds and three salt bridges (Figures 3P and 3Q).

5E3 epitope represents the major targets of gp42-specific antibody response in humans

Structurally, 6H2 binds to the juncture region of D-III and D-IV of gHgL (Figures 3B and 3C), away from the binding sites of AMMO1,²⁷ E1D1,³⁰ CL40,³¹ and 1D8³³ (Figure S5). The 10E4 epitope partially overlaps with that of AMMO1, but the majority of the epitope residues have never been reported as being recognized by other nAbs (Figure S5). These results suggest that four antibodies recognize four distinct epitopes and are distinguished from previously reported epitopes on the gHgL-gp42 ternary complex.

Next, we evaluated the immunodominance of these binding sites on the gHgL-gp42 complex using human sera from 20 EBV-positive healthy donors and 20 NPC patients in a competitive binding assay. Interestingly, those human sera could significantly reduce the binding of 5E3 to gp42 (healthy donors, 79.28% ± 4.33%; NPC patients, 77.00% ± 4.62%) (Figure 4A). On the contrary, we only observed weak competition (healthy donors, 18.10% ± 3.84%; NPC patients, 17.87% ± 4.05%) for 3E8 (Figure 4A), indicating that the epitope of 5E3, but not 3E8, contributed principally to induce anti-gp42 antibodies in humans.

Figure 4.

5E3, 6H2, and 10E4 interfere with EBV cell binding and cell fusion

(A and B) Blocking rates of sera from healthy EBV carriers and NPC patients were evaluated by competitive ELISA using anti-gp42 nAbs 3E8 and 5E3 (A) and anti-gHgL nAbs 6H2, 10E4, and 6H2+10E4 (B).

(C and D) Inhibition potency of anti-EBV nAbs against the binding of EBV glycoproteins to Akata B cells (C) and AGS epithelial cells (D). gHgL-gp42 complex (C) or gHgL (D) was pre-incubated with mAbs or PBS before being combined with Akata B cells (C) or AGS epithelial cells (D). Binding efficiency was measured using flow cytometry, and the mean fluorescence intensity (MFI) is shown. The antibody cocktail comprises equal ratios of the mAbs, with the total concentration of antibody administered equal for the antibody cocktail and single-mAb dosages. The positive control AMMO1 and negative control VRC01 were used.

(E and F) Inhibition potency of anti-EBV nAbs against mimic EBV membrane fusion of B cells (E) and epithelial cells (F).

(G) Blocking of anti-gp42 nAbs against the binding of HLA-II to gp42 immobilized on the chip, evaluated using SPR.

(H) Blocking of anti-gHgL nAbs against the binding of HLA-II to gHgL-gp42 complex immobilized on the chip, detected using SPR.

(I) Blocking of anti-gHgL nAbs against the binding of gHgL to EphA2 immobilized on the chip, detected using SPR.

(J) Structural comparison of the HLA-II binding site (colored yellow) and footprints of 5E3 (within the cyan line) and 6H2 (red line).

(K) Superimposition of the 6H2 variable domain and HLA-II (PDB: 1KG0) structures on the gHgL-gp42 complex shows steric clashing (indicated as a black dashed circle).

(L) Superposition of the structures of the 10E4 variable domain and EphA2-LBD (PDB: 7CZE) shows no epitope overlap or steric clashing.

(M) Structural comparison of the interactions of 10E4 to gL and receptor EphA2 to gL. The structures of 10E4:gL (khaki) and EphA2:gL (cyan; PDB: 7CZE) are superimposed. 10E4 binding induced a conformational change on the gL loop (aa 51–60), indicated by the black dashed arrow.

Data from one experimental replicate are shown in (A)–(I). All data are presented as the mean ± SEM from two or three replicates. Statistical analyses were performed using one-way ANOVA, comparing the data of each group with the negative control group (without adding antibody), and the result of the statistical test is shown above each column. One-way ANOVA was also performed to compare the individual antibody group and antibody cocktail group, and the result of the statistical test is shown above the lines. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001.

We also found that the sera in healthy EBV carriers blocked the binding of 6H2 to gHgL at a ratio of approximately 40%, while the sera from NPC patients showed a higher blocking efficiency (53.44% ± 9.52%) (Figure 4B). By contrast, the sera from both donors could only block the binding of anti-gHgL 10E4 at a lower level (23–26%) (Figure 4B). The serum competition efficacy against the combination binding of 10E4 and 6H2 could only be increased to 40.94% ± 9.17% for healthy donors and 45.99% ± 7.44% for NPC patients (Figure 4B). In summary, the human serum blocking results indicated that the 5E3 epitope is an immunodominant site for antibodies elicited by EBV infection in humans. Considering the excellent neutralizing and therapeutic efficacy of 5E3, its epitope may represent a vulnerable site for EBV infection of B cells.

Multiple mechanisms for the neutralization of anti-gHgL-gp42 antibodies

We next sought to explore the neutralizing mechanisms of the tested nAbs against different glycoproteins. The gHgL-gp42 complex is responsible for the binding of EBV to the receptor during B cell infection.⁴⁰ In the B cell binding assay, biotinylated gHgL-gp42 complex could effectively bind to Akata B cells (Figure 4C), mimicking viral attachment to host cells. Treatment of the gHgL-gp42 complex with 5E3, 6H2, or 10E4 alone could significantly reduce its binding to Akata B cells, whereas AMMO1 could only partially inhibit the binding, and VRC01 had a negligible effect (Figure 4C). The triple-antibody combination of 5E3, 6H2, and 10E4 (1:1:1 mixture) provided a more effective inhibition than either 6H2 or 10E4, whereas the difference between 5E3 alone and the antibody cocktail was unable to be determined because of limitation of detection (Figure 4C).

During epithelial cell infection, gHgL is responsible for binding directly to receptors.^17,18 Pre-incubation with 6H2, 10E4, or AMMO1, but not VRC01, had similar effects on blocking the binding of gHgL to AGS epithelial cells (Figure 4D). The combination of 6H2 and 10E4 showed enhanced inhibition of gHgL binding to the cell surface, suggesting a potential synergistic effect (Figure 4D). As expected, 5E3 had no effect on binding inhibition because gp42 does not participate in epithelial cell infection (Figure 4D).

Membrane fusion between the EBV envelope and the host cell membrane is triggered after binding of gHgL or the gHgL-gp42 complex to host cell receptors.¹¹ We further evaluated the abilities of nAbs to interfere with gHgL- or gHgL-gp42-mediated membrane fusion using a virus-free fusion assay, as described previously.³² In the B cell fusion assay, clear inhibition of cell fusion was observed in the presence of 5E3, 6H2, 10E4, or AMMO1, whereas VRC01 failed to elicit a response (Figure 4E). The triple-antibody combination could achieve better fusion inhibition efficiency to some extent (Figure 4E), although there was no significant difference between the antibody cocktail and any of the individual mAbs alone. In the epithelial cell fusion assay, the fusion activity could be significantly inhibited by the addition of 6H2, 10E4, or AMMO1 but not 5E3 or VRC01 (Figure 4F). In addition, the combination of 6H2 and 10E4 showed significant synergistic effects on the inhibition of epithelial cell fusion by blocking the fusion activity below detectable levels (Figure 4F).

The interaction between gp42 and the host cell receptor HLA-II is critical for EBV infection in B cells.⁴⁰ In accordance with previous reports,^27,40 we noticed that HLA-II could bind to the immobilized gp42 on the chip (Figure 4G). Pre-incubation of 5E3 with gp42 completely abrogated the binding of gp42 to HLA-II (Figure 4G), suggesting that 5E3 could significantly block the gp42:HLA-II interaction, which was consistent with the structural information showing a large overlapping region between the 5E3 epitope and the HLA-II binding site (Figure 4J). As a control, the binding signal between gp42 and HLA-II was unaltered by the addition of the non-neutralizing mAb 3E8, as expected (Figure 4G). Considering that anti-gHgL nAbs 6H2 and 10E4 could restrict the binding of the gHgL-gp42 complex to the B cell surface (Figure 4C), we further assessed their ability to influence the interaction between gHgL-gp42 and HLA-II. AMMO1, as a control, was incapable of restricting gHgL-gp42 binding to HLA-II (Figure 4H), consistent with a previous report.²⁷ Not surprisingly, the binding between HLA-II and gHgL-gp42 was unaffected by the presence of 10E4 (Figure 4H) because 10E4 binds to the distal D-I of gHgL, which is far away from the HLA-II binding site on gp42 (Figures 3B and 3C). While unexpected, 6H2 significantly interfered with the interaction between gHgL-gp42 and HLA-II (Figure 4H). Superimposing the structure of gp42:HLA-II⁴¹ onto our heptatomic complex structure showed obvious steric clashing between 6H2 and HLA-II (Figure 4K), indicating a different B cell-neutralizing mechanism of anti-gHgL nAbs.

Recent studies demonstrated that the interaction between gHgL and the cell receptor EphA2 is important for EBV entry into epithelial cells.¹⁸ A structural study revealed that the binding site of EphA2 is located mainly on the gL subunit of EBV.⁴² We found that 10E4 could interfere with the gHgL:EphA2 interaction, but such an effect could not be observed for 6H2, AMMO1, or VRC01 (Figure 4I), which was consistent with the findings of others.^27,32 Superimposition of structures of gHgL:EphA2 and gHgL:10E4 showed that there is neither epitope overlap nor steric clashing between bound 10E4 and EphA2 (Figure 4L). However, the binding of 10E4 would induce the obvious conformational changes that occur to the gL loop of aa 51–60 to overcome the steric clash between the bulky side chain of Q54 and the bound 10E4 (Figure 4M). The resultant conformation of the gL loop aa 51–60 region leads to further steric hindrance between Q54 and the bound EphA2 (Figure 4M). Such antibody-binding-induced conformational changes account for the blocking of receptor binding.

In summary, these data suggest that nAbs 5E3, 6H2, and 10E4 are capable of inhibiting the binding of the gHgL-gp42 complex to B cells and preventing B cell fusion. 5E3 and 6H2 could neutralize virus infection in B cells by blocking the gp42:HLA-II interaction. nAbs 6H2 and 10E4 can interfere with gHgL binding to epithelial cells, and 10E4 is also capable of inhibiting the gHgL:EphA2 interaction. The multiple mechanisms and diverse binding sites of these nAbs may together contribute to their in vitro synergistic neutralization of EBV infection.

Discussion

Here, we report that three nAbs recognize precisely three different regions of the gHgL-gp42 complex and, through in vitro and in vivo characterization, show that a cocktail of these three nAbs offers a synergistic inhibitory effect against EBV infection in vitro. Structural and functional analyses further highlight the value of understanding the gHgL-gp42 complex for the design of antibody cocktail therapeutics or vaccines against EBV infection.

These three nAbs—5E3, 6H2, and 10E4—offer discrete modes of neutralizing EBV infection: 5E3 blocks gp42 binding to HLA-II, in turn preventing gB-induced membrane fusion (Figures 4E, 5A and 5B). 6H2 creates a steric hindrance between the gHgL-gp42 complex and HLA-II (Figures 4K and 5C), which, in turn, accounted for its inhibition of B cell attachment and membrane fusion (Figures 4C and 4E). 6H2 also showed promising epithelial cell neutralization even though it did not affect the binding of gHgL to the known receptor EphA2¹⁸ (Figure 4I). We speculate that the binding of 6H2 may block gHgL from binding to some other unknown receptor or restrict the conformational change of gHgL required for subsequent gB activation to initiate membrane fusion with epithelial cells.³² A similar hypothesis was raised by others in the analysis of the nAb CL59 in its interaction with gHgL D-IV, with the authors speculating that EBV infection was neutralized by trapping gHgL in a nonfunctional state that fails to promote gB-mediated membrane fusion.³¹ Finally, the third nAb, 10E4, defined an epitope mainly on gL, which differs from the binding sites of the anti-gL nAb E1D1 and EphA2.^30,42 Indeed, in the absence of overlapping epitopes and steric clashing (Figure 4L), the binding of 10E4 restricts EphA2 binding by disrupting the conformation of gL (Figures 4I and 4M). In addition, 10E4 binding can induce loop motion that includes residue Q54, which has been reported previously as playing a critical role in the recruitment and activation of gB in B cell fusion;⁴³ our functional study indeed showed inhibition of cell fusion in B cells in the presence of this antibody (Figure 4E). We infer that 10E4 binding induces conformational changes that affect the gL-gB interaction required for gB-derived membrane fusion. In addition to the epitopes defined by 6H2 and 10E4, many other antigenic sites have been identified previously with the generation of anti-gHgL antibodies. For instance, E1D1, 770F8, and 769C2 mainly target D-I on gHgL,^30,44 similar to 10E4 in this study. The binding sites for other nAbs, such as 769B10, 796A7, AMMO1, and CL40, are located on D-I and D-II, which are key regions involved in gB activation.^31,43,44 1D8 and 769C5 recognize epitopes on the D-II and D-III regions.^33,44 Last, gHgL D-IV is also found to be targeted by CL59 and 770F7.^31,44 These findings indicate that gHgL is a critical target for many nAbs.

Figure 5.

Proposed neutralizing mechanisms of anti-gp42 and anti-gHgL nAbs

(A) Diagram of the interaction between the gHgL-gp42 complex and HLA-II during B cell infection, the interaction between gHgL and EphA2 during epithelial cell infection, and the conformational change to gB. Membrane fusion is mediated and promoted by this prefusion (Pre)-to-postfusion (Post) state change.

(B–D) Diagrams of the proposed mechanisms of neutralization of 5E3 (B), 6H2 (C), and 10E4 (D).

(B) 5E3 binds to an epitope that overlaps with the HLA-II binding site on gp42 and hence restricts the gHgL-gp42 complex from binding to the B cell surface by directly blocking the interaction between gp42 and HLA-II.

(C) 6H2 binds to the D-IV of gHgL and causes a steric clash that interferes with the binding of the gHgL-gp42 complex to HLA-II, thereby inhibiting subsequent membrane fusion.

(D) 10E4 binds to an epitope primarily on gL and inhibits gL binding to EphA2 by disrupting a key gL loop that may prohibit the interaction between gL and EphA2.

Created with BioRender.

As the indispensable component in B cell EBV infection,⁴⁰ gp42 is unique among all of the glycoproteins of human herpesviruses.¹³ However, there is a lack of neutralization strategies against dual-tropic EBV infection among B cells and epithelial cells, and this duality may have restricted the development of antibodies against gp42; indeed, compared with gp350, gB and gHgL, few anti-gp42 antibodies have been reported.^22,27,32,33 gp42 mainly consists of two domains: the N- and C-terminal domains.⁴⁰ The N-terminal domain is a continuous and flexible peptide wrapping around the exterior of three gHgL domains (D-II–D-IV) and plays a key role in regulating EBV tropism.¹⁹ A previous study highlighted that a panel of murine mAbs targeting the gp42 N-terminal domain was non-neutralizing,³⁶ suggesting that it may be difficult to induce the production of potent nAbs against the N terminus of gp42. However, given the important role of the gp42 C-terminal domain in receptor binding, this region may also serve as a major target for the development of anti-gp42 nAbs.³⁴ Indeed, 5E3, a gp42 C terminus-targeting antibody, has excellent B cell-neutralizing activity comparable with the anti-gHgL nAb AMMO1²⁷ (Figure 1C). The 5E3 nAb is directed to an epitope on gp42 that largely overlaps with the HLA-II binding site (Figure 4J); this is consistent with its receptor blocking activity because it inhibits B cell surface binding and the gp42:HLA-II interaction (Figures 4C and 4G). Importantly, our competition assays showed that the 5E3-like antibodies accounted for 77%–79% of the total anti-gp42-like antibodies produced in human sera from healthy individuals and NPC patients (Figure 4A), indicating that gp42 is likely to induce substantial 5E3-like nAbs during EBV infection in humans. In fact, previous studies have reported that the anti-gp42 nAb F-2-1 can efficiently neutralize B cell infection³⁷ by inhibiting gp42 binding to the receptor HLA-II on the cell surface.²⁰ Hence, our structural and functional data sufficiently support that the 5E3 epitope on gp42 may represent a key vulnerable site for EBV neutralization and a potential epitope for the production of valuable therapeutics.

Mühe et al.⁴⁵ compared the in vivo protection efficiency of those B cell-nAbs (BnAbs) and epithelial cell-nAbs (EnAbs) in a rhesus macaque animal model and demonstrated that high titers of BnAbs, but not EnAbs, could protect 1 one in 3 rhesus macaques against oral challenge of rhesus lymphocryptovirus. Together with our protection study of the anti-gp42 antibody in humanized mice, we propose that neutralizing B cell EBV infection may be a key and sufficient immune mechanism for sterilizing immunity against EBV infection. However, Mühe et al.⁴⁵ also noted that not all animals were afforded protection by BnAbs, indicating that EBV entry and pathogenesis are likely to be complicated.⁴⁵ Although no significant difference was observed between the protection efficacies of the antibody cocktail and 5E3 alone in our animal assay, the antibody cocktail does exhibit clear neutralizing synergy in vitro (Figures 1C and 1D). This may set the stage for future in vivo experiments, with adjustment of the dosages of the virus or antibody administration, to further explore the potential synergistic protection potency of antibody combination. Indeed, antibody cocktails have been widely studied and demonstrate clear value against viral infection, including COVID-19.^46,47 In the present study, our antibody cocktail showed synergistic neutralization in vitro, and the nAbs targeting different glycoproteins did not interfere with or negate the efficacy each other in the cocktail in vivo. Thus, despite the limitations of our study, we suggest that our antibody cocktail against multiple components of EBV may offer a strategy against EBV-related disease.

Previous vaccine studies of EBV have mainly focused on glycoprotein gp350, the most abundant glycoprotein on the EBV envelope, but studies have concluded that targeting gp350 alone is insufficient to provide complete protection against EBV infection.⁴⁸ A vaccine strategy that targets multiple viral components may be more beneficial in eliciting diverse nAbs against different glycoproteins and therefore provide better virus neutralization via multiple mechanisms. A recent study presented that the addition of gp42 into a gHgL-based subunit vaccine significantly improved the neutralizing activity and fusion inhibition efficiency of the immunized serum,³⁴ suggesting that gp42 is a promising addition for a multicomponent subunit vaccine against EBV. Furthermore, considering that 5E3-like anti-gp42 antibodies seem to be abundantly produced in the healthy population and NPC patients (Figure 4A), we surmise that the 5E3 epitope on gp42 is highly immunogenic. This information together suggests that gp42-based therapeutics and vaccine development should not be neglected for the control of EBV infection and disease.

Limitations of the study

Although the antibody cocktails exhibited in vitro synergistic neutralization, the advantage of the triple-antibody cocktail is not very obvious for in vivo protection. Considering the complexity of EBV infection in vivo, it requires well-designed experiments to determine a suitable viral dose and antibody usage to comprehensively compare the protection merit of mono-5E3 and the triple-antibody cocktail. Nevertheless, we believe that an antibody cocktail with comparable protection potency may have benefits in many aspects, such as dual-model neutralization, multiple neutralizing mechanisms, and probable resistance to viral escape.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
3E8	This paper	N/A
5E3	This paper	N/A
10E4	This paper	N/A
6H2	Hong et al.32	N/A
AMMO1	Snijder et al.27	N/A
VRC01	Zhou et al.38	N/A
MOUSE ANTI RABBIT CD4:FITC	Bio-Rad	Cat# MCA799F; RRID:AB_2075555
MOUSE ANTI RABBIT CD8:FITC	Bio-Rad	Cat# MCA1576F; RRID:AB_566891
MOUSE ANTI RABBIT T LYMPHOCYTES:FITC	Bio-Rad	Cat# MCA800F; RRID:AB_321388
MOUSE ANTI RABBIT IgM (B CELL MARKER)	Bio-Rad	Cat# MCA812GA; RRID:AB_10961295
Brilliant Violet 421™ Donkey anti-rabbit IgG (min. x-reactivity)	BioLegend	Cat# 406410; RRID:AB_10897810
Brilliant Violet 510™ anti-mouse CD45	BioLegend	Cat# 103138; RRID:AB_2563061
APC/Cyanine7 anti-human CD45	BioLegend	Cat# 304014; RRID:AB_314402
APC anti-human CD19	BioLegend	Cat# 392504; RRID:AB_2728416
FITC anti-human CD3	BioLegend	Cat# 344804; RRID:AB_2043993
Pacific Blue™ anti-human CD4	BioLegend	Cat# 317429, RRID:AB_1595438
PerCP/Cyanine5.5 anti-human CD8	BioLegend	Cat# 344710; RRID:AB_2044010
PE/Cyanine7 anti-human CD69	BioLegend	Cat# 310912; RRID:AB_314847
APC anti-human CD137 (4-1BB)	BioLegend	Cat# 309810; RRID:AB_830672
Brilliant Violet 650™ anti-human CD38	BioLegend	Cat# 356620; RRID:AB_2566233
PerCP/Cyanine5.5 anti-human CD24	BioLegend	Cat# 311116; RRID:AB_10960741
Rabbit anti-human CD20	Abcam	Cat# ab78237; RRID: AB_1640323
Goat Anti-Human IgG H&L (HRP)	Abcam	Cat# ab205718; RRID:AB_2819160
Alexa Fluor® 647 anti-human IgG Fc Antibody	BioLegend	Cat# 410713, RRID:AB_2728443
Goat anti-human IgG	Tianfun Xinqu Zhenglong Biochem.Lab	Cat# H0111-6
Bacterial and virus strains
EBV Akata GFP	Zhang et al.28	N/A
EBV CNE2 GFP	Zhang et al.28	N/A
Chemicals, peptides, and recombinant proteins
Ficoll-Paque PLUS	GE Healthcare	Cat# 17144003
LIVE/DEAD Aqua	Thermo Fisher	Cat# L34966
Complete Freund’s adjuvant	Sigma Aldrich	Cat# F5881
Incomplete Freund’s adjuvant	Sigma Aldrich	Cat# F5506
PrimeSTAR® GXL DNA Polymerase	Takara	Cat# DR050A
12-O-tetradecanoylphorbol 13-acetate	Beyotime	Cat# S1819
Sodium butyrate	Sigma Aldrich	Cat# V900464
Streptavidin -PE	eBioscience	Cat# 12-4317-87
Streptavidin APC	eBioscience	Cat# 17-4317-82
Red blood cell lysis buffer	Biolegend	Cat# 420301
Critical commercial assays
Ni Sepharose 6 Fast Flow	Cytiva	Cat#17-5318-03
EZ-Link™ Sulfo-NHS-LC-Biotin	Thermo Fisher Scientific	Cat# 21335
Superscript Ⅲ reverse transcriptase	Invitrogen	Cat# 18080093
Random hexamer	Invitrogen	Cat# N8080127
RNaseOUT Recombinant Ribonuclease Inhibitor	ThermoFisher Scientifc	Cat# 100000840
IGEPAL CA-720	Sigma Aldrich	Cat # 238589
Dual-Glo luciferase assay system	Promega	Cat# E2940
Tissue DNA kit	Omega	Cat# D3396-02
EBER detection kit	ZSGB-BIO	Cat# ISH-7001–100
Deposited data
CryoEM Maps of gp42:5E3:3E8 complex	This paper	EMD-33990
CryoEM Maps of gHgL:10E4 complex	This paper	EMD-33992
CryoEM Maps of gH:6H2 complex	This paper	EMD-33994
CryoEM Maps of gHgL-gp42:5E3:3E8:6H2:10E4 complex	This paper	EMD-33993
Atomic Model of gp42:5E3:3E8 complex	This paper	PDB 7YOY
Atomic Model of gHgL:10E4 complex	This paper	PDB 7YP1
Atomic Model of gH:6H2 complex	This paper	PDB 7YP2
Experimental models: Cell lines
CNE2-EBV	Zhang et al.49	N/A
293T	IMMOCELL	IM-H222
Akata-EBV	Molesworth et al.50	N/A
Akata	Wang et al.51	N/A
Daudi	ATCC	CCL-213
HNE1	Zhan et al.52	N/A
CHO-K1	ATCC	CCL-61
293F	Thermofisher	Cat# R79007
AGS	ATCC	CRL-1739
Experimental models: Organisms/strains
NOD.Cg-Prkdcem1IDMOIl2rgem2IDMO (NOD-PrkdcnullIL2Rγnull, NPI®)	BEIJING IDMO Co., Ltd	N/A
New Zealand White Rabbits	Songlian Laboratory Animal Center, Shanghai	N/A
Oligonucleotides
Forward primer specific for EBV BALF5 gene: 5′-GGTCACAATCTCCACGCTGA-3′	This paper	N/A
Reverse primer specific for EBV BALF5 gene: 5′-CAACGAGGCTGACCTGATCC-3′	This paper	N/A
Recombinant DNA
pCAGGS-gH	Hann et al.53	N/A
pCAGGS-gL	Hann et al.53	N/A
pCAGGS-gB	Hann et al.53	N/A
pCAG-T7	Okuma et al.54	N/A
pT7EMCluc	Okuma et al.54	N/A
pCDH-gp42	This paper	N/A
pCDNA3.1-EphA2	This paper	N/A
pCDNA3.1-DPA1	Dai et al.55	N/A
pCDNA3.1-DPB1	Dai et al.55	N/A
Software and algorithms
FlowJo software X 10.0.7	FlowJo	https://www.flowjo.com/
MotionCor2	Zheng et al.56	https://emcore.ucsf.edu/ucsf-software
Gctf	Zhang et al.57	https://en.wikibooks.org/w/index.php?title=Software_Tools_For_Molecular_Microscopy&stable=0#Gctf
cryoSPARC v3	Punjani et al.58	https://cryosparc.com
ResMap	Scheres et al.59	https://resmap.sourceforge.net/
Chimera	Pettersen et al.60	https://www.cgl.ucsf.edu/chimera/
Coot	Emsley et al.61	https://www2.mrc-lmb.cam.ac.uk/personal/pemsley/coot/
Molprobity	Chen et al.62	http://molprobity.biochem.duke.edu
Phenix	Adams et al.63	http://phenix-online.org
ChimeraX	Goddard et al.64	https://www.cgl.ucsf.edu/chimerax/
GraphPad Prism version 8	Graphpad	https://www.graphpad.com
Other
BD FACS Aria Ⅲ cell sorter	BD bioscience	https://www.bdbiosciences.com/en-us/products/instruments/flow-cytometers/research-cell-sorters/bd-facsaria-iii
LSRFortessa X-20 cytometer	BD Biosciences	https://www.bdbiosciences.com/en-us/products/instruments/flow-cytometers/research-cell-analyzers/bd-lsrfortessa-x-20
Biacore™ 8K	Cytiva	https://www.cytivalifesciences.com/en/us/shop/protein-analysis/spr-label-free-analysis/spr-systems/biacore-8k-p-05540
CytoFLEX	Beckman Coulter	https://www.beckman.com/flow-cytometry/research-flow-cytometers/cytoflex

Resource availability

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Yixin Chen (yxchen2008@xmu.edu.cn).

Materials availability

All requests for resources and reagents should be directed to the Lead Contact author. All reagents, which includes antibodies, proteins, plasmids, and virus, will be made available on request after completion of a Material Transfer Agreement for non-commercial usage.

Data and code availability

Structure coordinates have been deposited in the Protein DataBank under accession codes 7YOY (gp42:5E3:3E8), 7YP1 (gHgL:10E4) and 7YP2 (gH:6H2). The corresponding EM density maps have been deposited in the Electron Microscopy DataBank under accession numbers EMD-33990 (gp42:5E3:3E8), EMD-33992 (gHgL:10E4), EMD-33994 (gH:6H2) and EMD-33993 (gHgL-gp42:5E3:3E8:6H2:10E4). Reagents will be made available to the scientific community by contacting Lead contact author and completing a Material Transfer Agreement. This paper does not report original code. Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Experimental model and study participant details

Cell lines

All cell lines were incubated and routinely maintained at 37°C in humidified air containing 5% CO₂. CNE2 cells harboring a modified EBV,⁴⁹ AGS cells (ATCC) and 293T cells (IMMOCELL, Xiamen, Fujian, China) were grown in DMEM with 10% FBS. Akata cells harboring a modified EBV, EBV-negative Akata cells, Daudi cells and HNE1 cells were cultured in RPMI-1640 containing 10% FBS. CHO K-1 cells were maintained in Ham’s F-12 (Gibco) with 10% FBS. 293F cells (Thermo Fisher Scientific) were grown in SMM 293-TII Expression Medium (Sino Biological) with gentle shaking.

Human Specimens

Plasma samples from healthy individuals and NPC patients were provided by Sun Yat-sen University Cancer Center (Guangzhou, China). The use of human samples in this study was approved by the Institutional Ethics Committee of the Sun Yat-Sen University Cancer Center. Written informed consent and permissions were obtained from all participants.

Mice

All animal experiments were performed under protocols approved by the Sun Yat-sen University Cancer Center Animal Care and Use Committee. The humanized mice used in this study were the NOD-Prkdc^null IL2Rγ^null background. All animals were housed 5 per cage in a controlled environment in standard bedding with a standard 12-h daylight cycle, cessation of light at 6 p.m., and free access to standard chow diet and water. All mice used in this study were female and were 12 weeks old when experiments were initiated.

Method details

Plasmid

Detailed plasmid-construction of gHgL and gp42 has been previously published.^32,36 Briefly, the ectodomains of gH (aa 19–678), gL (aa 24–137) and gp42 (aa 34–223) were amplified from the EBV M81 strain (GenBank ID: KF373730.1). gL and gH were connected with a flexible linker. The ectodomain of EphA2 (aa 28–530) was obtained from the plasmid encoding the full-length of EphA2 (GenBank ID: NM_001329090.2) kindly provided by Mu-Sheng Zeng (Sun Yat-sen University Cancer Center). The genes encoding for gHgL, gp42 and EphA2 were cloned into pCDNA3.1 carrying an N-terminal CD5 signal peptide and a C-terminal 6×His tag. The construction of plasmids encoding a soluble form of HLA-DR2 was performed as previously described,⁵⁵ with some modifications. The genes of the DP2 alpha chain (DPA1∗0103) and the DP2 beta chain (DPB1∗0201) were synthesized by General Biosystems Co. Ltd (Chuzhou, China). The DP2 alpha chain was inserted into pCDNA3.1 with an N-terminal leader sequence, C-terminal basic zipper, and a FLAG tag. The DP2 beta chain was cloned into pCDNA3.1 with an N-terminal Aβb leader sequence followed by HLA DRα (aa 172–182), a C-terminal acidic zipper, and a His tag.

pCAGGS-gH, pCAGGS-gL, pCAGGS-gB, pCAG-T7, and pT7EMCLuc were kindly provided by Richard Longnecker (Northwestern University, USA). To construct the plasmid pCDH-gp42, the full-length sequence of gp42 was amplified from EBV-M81 BAC and inserted into the pCDH vector.

Recombinant expression

The 293F cells were transfected with plasmids encoding EBV glycoproteins, EphA2, HLA-II and recombinant antibodies using polyetherimide. At 7 days after transfection, the supernatant was collected and then filtered through 0.22-μm filters. Supernatants containing EBV glycoproteins or receptor proteins were purified using Ni²⁺ Sepharose 6 Fast Flow resin (Cytiva), followed by elution using an elution buffer (80 mM imidazole, 20 mM HEPES, 250 mM NaCl, pH 8.0). Supernatants containing recombinant antibodies were purified using Protein A affinity chromatography (GE Healthcare), and the antibodies were eluted using a different elution buffer (0.3 M glycine, pH 2.0 into 1 mL of 1 M Tris HCl, pH 8.0). Purified proteins and antibodies were then dialyzed into PBS, and verified by SDS-PAGE. gp42 and gHgL were further analyzed using a TSK-Gel 3000PWxl 30 cm × 7.8 mm column (TOSOH).

Rabbit immunization and mAb isolation

All animal experimental protocols were approved by the Xiamen University Laboratory Animal Management Ethics Committee. 10-week-old New Zealand White rabbits (Songlian Laboratory Animal Center) were immunized subcutaneously with 300 μg recombinant gp42 or gHgL mixed with an equal volume of Freund’s adjuvant (Sigma) three times at 2-week intervals. Antibody screening was performed as described previously.²⁸ Briefly, approximately 5 mL of blood was collected from immunized rabbits, and peripheral blood mononuclear cells (PBMCs) isolated using Ficoll-Paque PLUS (GE Healthcare), according to the manufacturer’s instructions. The gp42 and gHgL were biotinylated using Sulfo-NHS-LC-Biotin (Thermo Fisher Scientific), and separated PBMCs were incubated with biotinylated antigen at 4°C for 30 min. After three washes in PBS, PBMCs were further labeled with a panel of reagents: LIVE/DEAD Aqua (Thermo Fisher Scientific), CD4-FITC (Bio-Rad), CD8-FITC (Bio-Rad), T Lymphocyte-FITC (Bio-Rad), IgM-RPE (Bio-Rad), APC streptavidin (BioLegend). B cell sorting was performed using a FACS Aria III Sorter (BD Biosciences). Antigen-specific B cells were sorted directly into 96-well plates containing lysis buffer. cDNA was obtained through reverse transcription from extracted RNA of positive B cells using Superscript Ⅲ reverse transcriptase (Invitrogen) primed with random hexamers. Antibody variable regions of the heavy and light chains were recovered via two rounds of PCR using GXL polymerase (Takara), and then cloned into pVRC8400 vector containing the heavy-chain and light-chain constant regions of rabbit IgG subtype antibody.

Enzyme-linked immunosorbent assay (ELISA)

gp42 or gHgL protein (100 ng/well) was coated into the wells of 96-well microplates at 37°C for 2 h, prepared in 0.1 M NaHCO₃/Na₂CO₃ (pH 9.6). Plates were washed once with PBS containing 0.1% v/v Tween 20 (PBST) and then blocked with blocking solution (PBS containing 2% w/v non-fat dry milk) for 2 h at 37°C. A series of human sera were added to the blocked microplates and the plates incubated at 37°C for 30 min. After five washes with PBST, each well was incubated with 100 μL of diluted horseradish peroxidase (HRP)-conjugated goat anti-human IgG (Abcam) for 30 min at 37°C. After five washes with PBST, color development was induced by the addition of 100 μL tetramethylbenzidine substrate (Wantai BioPharm) for 15 min in the dark at 37°C. The reaction was quenched with 50 μL 2 M H₂SO₄, and absorbance read at OD_{450 nm/630 nm} on a PHOmo microplate reader (Autobio).

Competitive binding between human sera and antibodies were performed using ELISA assay. Briefly, concentrations of diluted human sera were measured to identify a concentration that would provide an OD of ∼1 using ELISA. In the blocking assays, diluted human sera were added into gp42-or gHgL-coated ELISA plates, respectively, and incubated at 37°C for 30 min. After five washes, the diluted HRP-conjugated antibodies were added at the concentration of their EC₅₀, and the plates incubated for another 30 min. After five washes with PBST, color development was performed as described for the indirect ELISA. The percent inhibition of human sera by antibodies was calculated as: 1−([OD value of the antibodies binding to coated glycoproteins pre-incubated with human sera]/[OD value of the antibodies binding to coated glycoprotein without human sera treatment]) × 100%.

Surface plasmon resonance assay (SPR)

Antibody binding kinetics against gp42 or gHgL were measured by SPR on a BIAcore 8K (Cytiva). To analyze the equilibrium dissociation constant (K_D) of the antibodies, 2 μg/mL antibody was captured by a Sensor Chip Protein A at a flow rate of 10 μL/min for 60 s. Serially diluted gp42 or gHgL was injected at a flow rate of 30 μL/min in PBS-P buffer (Cytiva) for 120 s, followed by a period of dissociation at 30 μL/min for 120 s. BIAcore Insight Evaluation software was used to analyze the results, and a 1:1 binding model was used for curve fitting.

To analyze epitope competition for anti-gp42 or anti-gHgL mAbs, gp42 and gHgL were covalently immobilized onto a Sensor Chip CM5 (Cytiva), respectively, and injected with corresponding mAbs at 300 nM for 300 s successively. To assess whether the mAbs could simultaneously bind to the gHgL-gp42 complex, the abovementioned gp42-immobilized chip was injected with a solution of 100 nM gHgL. Then 300 nM anti-gp42 mAbs and anti-gHgL mAbs were loaded for 300 s successively.

To analyze competitive binding to gp42 among anti-gp42 mAbs and HLA-II, gp42 was covalently immobilized onto a Sensor Chip CM5 (Cytiva). The chip was then injected with running buffer, 3E8 (300 nM) and 5E3 (300 nM) for 300 s, respectively, followed by loading of HLA-II (300 nM) for 60 s.

To analyze potential competitive binding between anti-gHgL mAbs and HLA-II to the gHgL-gp42 complex, the abovementioned gp42-immobilized chip was injected with 100 nM gHgL for 300 s followed by solutions of running buffer or 300 nM 6H2, 10E4 and AMMO1 for 300 s, respectively. Subsequently, the chip was loaded with a solution of 300 nM HLA-II for 60 s.

To analyze the competitive binding to gHgL of anti-gHgL mAbs and EphA2, EphA2 was immobilized onto a Sensor Chip CM5 (Cytiva). gHgL (1,000 nM) was preincubated with equal concentrations of corresponding antibodies or protein (6H2, 10E4 and AMMO1) for 30 min at 37°C, then injected onto the EphA2-immobilized chip for 60 s.

Virus production

To produce epithelial cell-tropic virus, Akata-EBV-GFP cells were suspended at 2 × 10⁶ cells/mL in RPMI 1640. Goat anti-human IgG (Promega) was added to a final concentration of 100 μg/mL and incubated at 37°C for 6 h to induce EBV production. B cell-tropic virus was produced from CNE2-EBV-GFP cells. The cells were seeded in 10-cm cell plates at 90% confluence, and 12-O-tetradecanoylphorbol 13-acetate (TPA) (20 ng/mL) and sodium butyrate (2.5 mM) were added to induce virus production.

After induction, the culture media were changed with fresh RPMI 1640 containing 10% FBS for another 72 h. Subsequently, the supernatant was harvested and passed through a 0.45-μm filter. To further concentrate virus, the supernatant was centrifuged at 50,000 × g for 2.5 h, and resuspended in FBS-free RPMI 1640. Virus was stored at −80°C for use.

Binding to induced CNE2-EBV-GFP cells

CNE2-EBV-GFP cells were induced as described above, and further digested and harvested. Cells were incubated with human FcR blocking reagent for 30 min at 4°C. followed by 1 μg of 3E8, 5E3, 6H2, 10E4, AMMO1 or VRC01, and incubated at 4°C for a further 30 min. Cells were washed three times with PBS and then stained with goat anti-human IgG Alexa Fluor 647 (BioLegend) at 1:500 dilution. Data were collected using an LSRFortessa X-20 cytometer (BD Biosciences), and further analyzed by FlowJo software X 10.0.7 (Tree Star).

Neutralization assay

For B cell neutralization, antibodies (starting from 100 μg/mL) were serially diluted in duplicate wells of 96-well plates containing 50 μL RPMI 1640. CNE2-EBV-GFP (2,000 GRUs; 20 μL) was added to the plates and the reactions incubated at 37°C for 2 h. Wells were then combined with 130 μL of fresh 10% FBS RPMI 1640 containing 10⁴ Akata B cells and further incubated at 37°C for 48 h.

For epithelial cell neutralization, serially diluted antibodies (starting from 100 μg/mL) were incubated with 50 μL (2,500 GRUs) Akata-EBV-GFP for 2 h at room temperature. The mixture was added to 5 × 10³ HNE1 epithelial cells/well in 96-well plates and then incubated for 3 h at 37°C. The mixture was then aspirated and replaced with fresh medium.

Rates of infection were measured by detecting the numbers of GFP-positive cells on an LSRFortessa X-20 cytometer (BD Biosciences). Uninfected cells were used as negative controls, and cells infected by EBV without antibody treatment were performed as positive controls. FlowJo software X v10.0.7 (Tree Star) was used to analyze the number of GFP-positive cells. Neutralization (%) was defined as: 1 – (% infected cells in the antibody-added well)/% of infected cells in the positive control well with virus alone) × 100%.

The inhibition profiles (IC₅₀ values) were fitted using the log (inhibitor) versus response – variable slope (four parameters) analysis in GraphPad Prism 8 software.

EBV infection in humanized mice

4-week-old NOD.Cg-Prkdc^em1IDMOIl2rg^em2IDMO (NOD-Prkdc^null IL2Rγ^null, NPI) mice (BEIJING IDMO Co., Ltd) were administered with an intraperitoneal (i.p.) injection of a single dose of Busulfan. After 48 h, the mice were grafted intravenously (i.v.) with human cord blood-derived CD34⁺ hematopoietic stem cells to construct the humanized mice. After 8 weeks post-grafting, humanized mice were i.p. injected with a single experimental antibody, a cocktail of antibodies, or the control antibody at a dose of 20 mg/kg body weight. After 24 h, all mice were challenged with a dose of 25,000 GRUs Akata-EBV-GFP via i.v. injection. In the following weeks, peripheral blood samples were collected. Body weight changes were tracked weekly. Mice were euthanized at 10 weeks’ post-challenge or earlier if they showed signs of being clinically ill (e.g., body weight loss of approximately 20%).

Detection of EBV DNA in blood and tissues

DNA was extracted from peripheral blood samples (50 μL) and splenocytes of mice following the instructions provided with the commercial DNA extraction kit (Omega). A fragment of EBV BALF5 gene was quantified by real-time PCR (RT-PCR) with the following primers: F: 5′-GGTCACAATCTCCACGCTGA-3’; R: 5′-CAACGAGGCTGACCTGATCC-3’.

Immunophenotype analysis of human cells in humanized mice

Peripheral blood samples and splenocytes from mice were treated with commercial red blood cell lysis buffer (BioLegend) at room temperature for 10 min. Cells were then centrifuged at 300 × g for 5 min, washed twice with PBS, resuspended in PBS, and stained with antibodies, including anti-human CD45-APC/Cy7, CD19-APC, CD3-FITC, CD4-pacific blue, CD8-PC5.5, CD137-APC, CD69-PC7, CD24-PC5.5, CD38-BV650 and anti-mouse CD45-BV510 (BioLegend) for 30 min at 4°C. Flow cytometry assays were performed with a CytoFLEX (Beckman Coulter), with the data analyzed using FlowJo software X v10.0.7 (Tree Star).

H&E staining, IHC, and in situ hybridization

Tissue samples were fixed in 10% formalin and embedded in paraffin, according to standard laboratory protocols. Samples were stained with hematoxylin and eosin (H&E). Human B cell immunostaining was performed using anti-hCD20 antibody (Abcam) at a 1:200 dilution. EBV-encoded RNAs (EBERs) were detected by in situ hybridization with an EBER detection kit (ZSGB-BIO), according to the manufacturer’s instructions.

Cryo-EM sample preparation and data collection

Purified gHgL-gp42:5E3:3E8:6H2:10E4 complex was diluted to a concentration of 3.0 mg/mL in PBS containing 0.25 mM n-Dodecyl-β-D-Maltopyranoside (DDM, Sigma). Sample aliquots of 3 μL were loaded onto holey carbon Quantifoil grids (R1.2/1.3, 200 mesh, Quantifoil Micro Tools) that were glow-discharged at 20 mA for 80 s, and then vitrified with a Vitrobot Mark IV (Thermo Fisher Scientific) with a blotting time of 5 s at 100% humidity and 4°C. Data were acquired on a Tecnai F30 transmission electron microscope (Thermo Fisher Scientific) operating at 300 kV and equipped with a Gatan K3 direct electron detector. Images were collected using the SerialEM software at a nominal magnification of 39,000× at super-resolution mode, corresponding to a pixel size of 0.389 Å. The total electron dose was set to 60 e⁻ Å⁻² subdivided into 36 frames over a 4.5-s exposure.

Image processing and 3D reconstruction

All datasets were processed with cryoSPARC v3.⁵⁸ Raw movie stacks were motion-corrected using MotionCor2.⁵⁶ Contrast transfer function (CTF) parameters were estimated using Gctf.⁵⁷ Low-quality images were discarded before reconstruction. The full reconstruction procedures are summarized in Figure S4B. In brief, particles were automatically picked using the “Blob picker” or “Template picker,” and selected through several rounds of reference-free 2D classifications. Selected good particles were then subjected to ab-initio reconstruction, heterogeneous refinement, and non-uniform refinement. Due to flexibility at the antibody interface, localized refinements of regions for 3E8 and 5E3 bound to gp42, 10E4 bound to gHgL and 6H2 bound to gH were performed by specific local masks to improve the quality of cryo-EM density maps. These maps of localized refinement were used for further model building. The final resolution was determined by gold-standard Fourier shell correlation (FSC) between the two independently refined half maps, with a cutoff of 0.143.⁵⁹ Local map resolution was estimated with ResMap.⁶⁵

Model building, refinement, and 3D visualization

Initial models were generated using homology modeling by Accelrys Discovery Studio software (available from: URL: https://www.3dsbiovia.com). The gHgL-gp42-E1D1 (PDB: 5T1D) structure was used as the initial model for gHgL-gp42. Initial templates were fitted into the corresponding final cryo-EM maps using Chimera,⁶⁰ and further corrected and adjusted manually by real-space refinement in Coot.⁶¹ The resulting models were then refined using phenix.real_space_refine in Phenix.⁶³ Models were subjected to iterative cycles of manual model adjustment using Coot and Phenix. The final atomic models were validated using Molprobity.^62,66 Cryo-EM data processing and refinement statistics are summarized in Table S1. The buried surface areas and interactions were analyzed using PISA (https://www.ebi.ac.uk/pdbe/pisa/) and the CCP4 program.⁶⁷ Structural representations were generated using Chimera or ChimeraX.^64,68

Cell-surface binding assay

For cell-surface binding assays, gHgL was biotinylated using an EZ-Link Sulfo-NHS-LC-Biotin reagent (Thermo Fisher Scientific). Biotinylated gHgL (1 μg) conjugated with SA-PE (gHgL-PE) was diluted in 10 μL PBS and added to individual wells of a 96-well plate. 3E8, 5E3, 6H2, 10E4, AMMO1, VRC01 and antibody combinations (7 μg for B cells, 2 μg for epithelial cells) were added to the 96-well plate containing gHgL-PE with (B cell) or without gp42 (epithelial cell) and then incubated at room temperature for 1.5 h. Meanwhile, adherent AGS cells were trypsinized, washed with PBS, and then incubated in RPMI 1640 at 37°C for 30 min to recover. Akata and recovered AGS cells were centrifuged at 300 × g for 5 min and resuspended in ice-cold 1% BSA (Sigma) in PBS. 100 μL of Akata and AGS cells at a density of 1 × 10⁶ cells/mL were mixed with gHgL-PE with/without gp42 and antibodies, and incubated on ice for 1 h. Cells were collected by centrifugation at 300 × g for 5 min, washed three times and resuspended with ice-cold 1% BSA in PBS. The proportion of PE-stained positive cells was determined on an LSRFortessa X-20 cytometer (BD Biosciences). Data were analyzed, and the mean fluorescence intensity (MFI) of PE calculated using FlowJo software X v10.0.7 (Tree Star).

Virus-free fusion inhibition assay

For epithelial cell fusion, 293T and CHO-K1 cells were seeded into 10-cm dishes and grown to 80% confluence. Effector CHO-K1 cells were transfected with plasmids encoding gH, gL, gB, and T7 polymerase (2.5 μg for each plasmid), while target 293T cells were transfected with 10 μg of pT7EMCLuc that expresses the luciferase under the control of T7 promoter. 24 h after transfection, 2.5 μg antibody per test were added to trypsinized effector cells (2 × 10⁵ cells/test), and incubated at 37°C for 30 min. Approximately 2 × 10⁵ target 293T cells were then mixed with aforementioned effector cells, and further cultured in 24-well plates. After 24 h, the culture media were aspirated and the cells were lysed with 100 μL of Dual-Glo luciferase reagent (Promega). Next, 80 μL of lysate was added to white-bottom plates to measure luciferase activity.

For B cell fusion, effector CHO-K1 cells were transfected with plasmids encoding for gp42, gH, gL, gB, and T7 polymerase (2 μg for each plasmid). Target Daudi B cells were transfected with pT7EMCLuc. The rest of the process was performed as described above.

Quantification and statistical analysis

Unless noted otherwise, one-way ANOVA was used to assess statistical significance. All statistical analyses were conducted with GraphPad Prism version 8. The number of replicates and a description of the statistical method are provided in the corresponding figure legends. p Values of <0.05 were considered statistically significant. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001, ns = not significant.

Published: November 21, 2023